Explore data-driven approaches for memory architecture design in modern datacenters through this insightful conference talk. Delve into the critical role of memory in system performance and cost, challenging traditional decision-making methods in computer architecture. Examine two system-level memory problems: Exascale supercomputing and Cloud memory disaggregation. Learn how quantitative and analytical approaches can guide memory architecture decisions, moving beyond "rules of thumb" and "zeitgeist." The talk, presented by Dan Ernst at the SPCL_Bcast event, includes an introduction, main presentation, brief announcement, and an engaging Q&A session, offering a comprehensive look at memory-centric designs for cutting-edge datacenter technologies.
Follow the Data: Memory-Centric Designs for Modern Datacenters
Scalable Parallel Computing Lab, SPCL @ ETH Zurich via YouTube
Overview
Syllabus
Introduction
Talk
Announcement
Q&A Session
Taught by
Scalable Parallel Computing Lab, SPCL @ ETH Zurich